Association of prescribed opioid analgesics, benzodiazepines, and antidepressants with duration of hospitalization: A national cohort study

Treatment of chronic non-cancer pain (CNCP) with opioid analgesics (OAs) has burgeoned along with concomitant prescribing of benzodiazepines (benzos) and antidepressants (antideps). It is unknown if these drugs are associated with duration of hospitalization for any cause. For this analysis, we identified persons filling ≥2 prescriptions (Rxes) for non-injectable Schedule II or III OAs for CNCP from enrollees aged 18 to 64 in Aetna's national HMO (2.1 million) from 1/2009 to 7/2012. Of 279,049 such persons aged 18-64 and enrolled ≥1 year, we excluded 57,986 persons with: cancer diagnosis (not basal cell) within 6 months of an OA Rx, OAs for opioid dependence, or incomplete diagnostic or demographic data. Claims and enrollment data files offer: demographics, all encounters, ICD-9-CM codes, and filled Rxes. From ICD-9 codes, we created indicators for 5 categories of CNCP, depression, and anxiety disorder. In each 6-month interval after each patient's first OA Rx until last enrollment or July 2012, we categorized OA risk from the dose and duration of all prescribed OAs: NO risk (0 days), LOW risk (1-59d at <40 morphine equivalent daily dose in mg [MED)], MEDIUM risk (1-59d at 40-100 MED or ≥60d at <40 MED), HIGH risk (>100 MED or ≥60d at 40-100 MED). We calculated duration of benzo Rxes (0d, 1-30d, 31-90d, and 91-180d) and antidep Rxes (0d, 1-60d, and 61-180d) in these 6-month intervals. Using repeated measures Poisson regression, we examined the association of inpatient days per person per 6-months with OA risk and interactions with benzos and antideps adjusting for demographics, health care, and clinical variables. Our cohort totaled 221,063 with median age 45 (IQR=35-54), 57% female, and U.S residence in South (47%), West (18%), Midwest (6%) and Northeast (29%). Of 881,555 six-month intervals, 6% were hospitalized for median of 3 days (IQR=2-6). After adjustment, inpatient days rose significantly (p<0.001) with OA risk vs. NO risk (HIGH: incident rate ratio [IRR] [95% CI]=3.27[2.91,3.67], MEDIUM: IRR=2.19[1.93,2.49], LOW: IRR=1.38[1.17,1.64]). OA risk interacted significantly with benzos and antideps, respectively (p<0.001). Compared with NO OA risk + no benzos, HIGH OA Risk + benzos for 1-30d had over a 5-fold increase in length of stay (LOS) (IRR=5.48[4.82,6.23]). Compared with NO OA risk + no antideps, HIGH OA risk + antideps for 1-60d had a nearly 5-fold increase in LOS (IRR=4.53[3.86,5.31]). In persons with CNCP, higher risk OAs along with short-term use of benzos or antideps was associated with more inpatient days.